Tool Directory / Monitoring
Know when production breaks before users tell you
Monitoring is the layer teams add after the incident that made them wish they had it. The good news is that the minimum viable setup is genuinely small: error tracking, product analytics, and an uptime check. Everything past that is optimization.
Reviewed August 2026. Back to the tool directory.
What this layer decides
Four questions, and they want four different tools. Did something throw? That is error tracking. Are people using the feature? That is product analytics. Is the site up right now? That is uptime monitoring. Why is it slow? That is tracing and metrics, and it is the only one you can safely defer.
Two traps. Alert fatigue kills a monitoring setup faster than having no monitoring at all: if a notification does not require action, it should not page anyone. And every client-side tool adds JavaScript to your pages, so budget the weight and check the privacy implications before installing three of them.
Error tracking and product analytics
Sentry
Free tier, Team from about 26 dollars
The reference error tracker, and still the best at the core job: grouping identical exceptions, mapping minified stack traces back to source, and telling you which release introduced a regression. The free developer plan covers roughly 5,000 errors a month plus a small allowance of replays, spans, and monitors. Performance tracing and session replay are bolted on well but priced separately. sentry.io
PostHog
1M events free, usage priced after
Product analytics, session replay, feature flags, A/B testing, surveys, and error tracking in one platform, which removes a lot of glue work and several separate invoices. The free tier is unusually generous - about a million events, thousands of replays, and a large error allowance every month. Open source and self-hostable if data residency matters. posthog.com
Microsoft Clarity
Free, no volume cap
Heatmaps, session recordings, and rage-click detection at no cost and with no traffic limit, which is a genuinely unusual offer. It answers "what are people actually clicking?" better than any event dashboard. It is not an analytics warehouse and cannot replace one, but as a free second opinion on a confusing funnel it is hard to argue with. clarity.microsoft.com
Infrastructure metrics and uptime
Grafana
The dashboard layer over Prometheus for metrics, Loki for logs, and Tempo for traces. Grafana Cloud has a real free tier and self-hosting is free. Powerful and genuinely deep, but each backend has its own query language - budget the learning time honestly.
OpenTelemetry
Not a product - the vendor-neutral standard for emitting traces, metrics, and logs. Instrument with OTel and you can change backends by editing an exporter config instead of ripping out an SDK. Worth doing on day one purely as insurance.
Better Stack
Uptime checks down to 30-second intervals, phone and push alerting, on-call schedules with escalation, and a hosted status page, plus log management in the same product. The pragmatic choice when you want incident response without assembling it.
UptimeRobot
A free plan that monitors a useful number of endpoints and emails you when one stops answering. No on-call rotation, no escalation policy, nothing clever. For a side project or a small site that is the entire requirement.
Our pick
Sentry plus PostHog, and an uptime check that pages you
Install Sentry for errors and PostHog for product analytics. Sentry is still meaningfully better at the specific job of grouping exceptions and pointing at the offending release, and PostHog covers analytics, replay, and feature flags on one generous free tier instead of three separate subscriptions. Add Microsoft Clarity when you want free heatmaps as a second opinion.
Then add one uptime check that actually reaches a human - UptimeRobot for a side project, Better Stack once someone is on call. Instrument with OpenTelemetry from the start so the backend stays swappable, and defer Grafana and full tracing until you have a performance question that logs cannot answer. Every alert should demand an action, or it will train your team to ignore all of them.
Keep going
Hosting layer
Some platforms ship half of this already. Check before you buy.
Browse the layerTesting layer
Catch it before deploy so monitoring has less to report.
Browse the layerSaaS stack guide
Observability in context with the rest of a production product.
See the stackHTTP status codes
Know which codes belong on a dashboard and which are noise.
Grab the cheatsheetRunning an AI feature? Token spend belongs on the same dashboard as errors - the AI layer covers what to track, and the internal tools stack shows how much of this you can skip behind a firewall.